Football Players Analysis
A Data Science Project by:
PRASFUR TIWARI & RAVI SISTA
Data Analysis¶
In [4]:
# import libraries
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')
plt.style.use('ggplot')
In [5]:
# importing the data set
df = pd.read_csv('C:\\Users\\smile\\Desktop\\Prasfur\\Project 1 - FIFA\\Data\\FIFA 19.csv')
In [6]:
# exploring the data set
df.head()
Out[6]:
In [7]:
df.shape
Out[7]:
In [8]:
# Listing out the columns
df.columns
Out[8]:
In [9]:
# Remove unwanted columns
df.drop(['Unnamed: 0'],axis=1,inplace=True)
In [10]:
# Listing info of the data set
df.info()
In [11]:
# Checking the number of missing values
df.isnull().sum()
Out[11]:
In [12]:
#Checking extent of null values
sns.heatmap(df.isnull(),yticklabels=False)
Out[12]:
In [13]:
# Since the column loaned from almost has no values, we will drop it
df.drop(['Loaned From'],axis = 1, inplace=True)
In [14]:
#now the data which have NA values, we will fill them with the mean value of that column
df.fillna(df.mean(),inplace=True)
In [15]:
df.isnull().sum()
Out[15]:
In [16]:
#Checking extent of null values
sns.heatmap(df.isnull(),yticklabels=False)
Out[16]:
In [17]:
#there are still cells in which the mean value could not be assigned. This may be because those columns have strings. So we will assign a value "Unassigned" to the dataset
df.fillna("Unassigned",inplace=True)
In [18]:
df.isnull().sum()
Out[18]:
In [19]:
df.dtypes
Out[19]:
In [20]:
#Displaying all columns
df.keys()
Out[20]:
In [21]:
# Final data clean-up procedure
df.drop(['Photo','Flag','Club Logo','Real Face','Special'],axis=1,inplace=True)
df.head()
Out[21]:
In [22]:
df.isnull()
Out[22]:
In [23]:
df.isnull().sum()
Out[23]:
In [24]:
#Checking extent of null values
sns.heatmap(df.isnull(),yticklabels=False)
Out[24]:
In [25]:
# Getting insights from dataset
df.describe()
Out[25]:
Data Visualization¶
In [26]:
# Age distribution of players
# Histogram: number of players's age
sns.set(style ="dark", palette="colorblind", color_codes=True)
x = df.Age
plt.figure(figsize=(12,8))
ax = sns.distplot(x, bins = 58, kde = False, color='g')
ax.set_xlabel(xlabel="Player\'s age", fontsize=16)
ax.set_ylabel(ylabel='Number of players', fontsize=16)
ax.set_title(label='Histogram of players age', fontsize=20)
plt.show()
oldest = df.loc[df['Age'].idxmax()]
print("The oldest player in FIFA 19 is", df['Age'].max(), "years old. His name is", oldest['Name'],
'he is from',oldest['Nationality'],'and plays for',oldest['Club'],'.')
print('The median age of a player on FIFA 19 is', np.mean(df['Age']))
youngest = df.loc[df['Age'].idxmin()]
print('The youngest players is',df['Age'].min(), "years old. His name is", youngest['Name'],
'he is from',youngest['Nationality'],'and plays for',youngest['Club'],'.')
In [27]:
# Overall rating distribution of players
plt.hist(df['Overall'])
plt.xlabel('Players Rating ')
plt.ylabel('Number of players')
plt.show()
best = df.loc[df['Overall'].idxmax()]
print("The best player in FIFA 19 is", df['Overall'].max(), "overall. His name is", best['Name'],
'he is from',best['Nationality'],'and plays for',best['Club'],'.')
print('The median rating of a player on FIFA 19 is', np.mean(df['Overall']))
worst = df.loc[df['Overall'].idxmin()]
print('The worst players is',df['Overall'].min(), "overall. His name is", worst['Name'],
'he is from',worst['Nationality'],'and plays for',worst['Club'],'.')
In [28]:
# Distribution of Player's potential rating
plt.hist(df['Potential'], color = 'blue')
plt.xlabel('Players Potential')
plt.ylabel('Number of players')
plt.show()
bestp = df.loc[df['Potential'].idxmax()]
print("The best potential player in FIFA 19 is", df['Potential'].max(), "overall. His name is", bestp['Name'],
'he is from',bestp['Nationality'],'and plays for',bestp['Club'],'.')
print('The median potential rating of a player on FIFA 19 is', np.mean(df['Potential']))
worstp = df.loc[df['Potential'].idxmin()]
print('The worst potential player is',df['Potential'].min(), "overall. His name is", worstp['Name'],
'he is from',worstp['Nationality'],'and plays for',worstp['Club'],'.')
In [29]:
# Displaying the Age vs Overall performance
plt.figure(figsize=(10, 5))
sns.regplot(df['Age'] , df['Overall'])
plt.title('Age vs Overall rating')
plt.show()
In [30]:
# Displaying the Age vs Potential performance
plt.figure(figsize=(10, 5))
sns.regplot(df['Age'],df['Potential'],color = 'green')
plt.title('Age vs Potential rating')
plt.show()
In [31]:
# Plotting the heatmap of the data set
plt.figure(figsize=(10,8))
sns.heatmap(df.corr(),linewidths=3)
plt.title('Dataset Heatmap')
plt.show()
In [32]:
# Preferred foot analysis
plt.figure(figsize=(5,5))
sns.countplot(df['Preferred Foot'])
Out[32]:
In [33]:
# Displaying the Top 25 players
df_best_players = pd.DataFrame.copy(df.sort_values(by = 'Overall' ,
ascending = False ).head(25))
plt.figure(figsize=(20, 10))
plt.bar('Name' , 'Overall' , data = df_best_players, width=0.5, color = 'Purple')
plt.xlabel('Players names', fontsize=30)
plt.xticks(rotation = 90,fontsize=20, fontname='monospace')
plt.ylabel('Overall Rating', fontsize=30)
plt.title('Top 25 players Overall Rating', fontsize=40)
plt.ylim(87 , 95)
plt.show()
In [34]:
#Stamina vs Sprint Speed Plot
df.plot(kind = 'scatter' , x='Stamina' , y = 'SprintSpeed' , alpha = .5 )
plt.xlabel('Stamina')
plt.ylabel('Sprint Speed')
plt.title('Stamina-Sprint Speed Scatter Plot')
plt.show()
In [35]:
# Displaying the relation between Age & Sprint Speed
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['SprintSpeed'], color = 'Green')
plt.title('Age vs Sprint Speed')
plt.show()
In [36]:
# Displaying the relation between Overall & Potential
df.groupby('Overall')['Potential'].mean().plot()
plt.title('Overall vs Potential')
plt.ylabel("Potential",rotation=90)
plt.show()
In [37]:
#count of players Position Wise
plt.figure(1,figsize=(20,10))
p = sns.countplot(x = 'Position', data = df,palette='inferno_r')
p.set_title(label='Count of Players', fontsize=25)
Out[37]:
In [38]:
#Cleaning some of values so that we can interpret them
def value_to_int(df_value):
try:
value = float(df_value[1:-1])
suffix = df_value[-1:]
if suffix == 'M':
value = value * 1000000
elif suffix == 'K':
value = value * 1000
except ValueError:
value = 0
return value
df['Value'] = df['Value'].apply(value_to_int)
df['Wage'] = df['Wage'].apply(value_to_int)
In [39]:
# Displaying the Values of players
df1 = pd.DataFrame.copy(df.sort_values(by = 'Value' , ascending = False ).head(25))
plt.figure(figsize=(20, 5))
plt.bar('Name' , 'Value' , data = df1, width=0.5, color = 'Orange')
plt.xlabel('Players names', fontsize=30)
plt.xticks(rotation = 90,fontsize=20, fontname='monospace')
plt.ylabel('Value', fontsize=30)
plt.title('Value of Players(in Millions)', fontsize=30)
plt.show()
In [40]:
# Distribution of Wage of Players
df2 = pd.DataFrame.copy(df.sort_values(by = 'Wage' , ascending = False ).head(25))
plt.figure(figsize=(20, 5))
plt.bar('Name' , 'Wage' , data = df2, width=0.5, color = 'Red')
plt.xlabel('Players names', fontsize=30)
plt.xticks(rotation = 90,fontsize=20, fontname='monospace')
plt.ylabel('Wage', fontsize=30)
plt.title('Wage of Players(in Millions)', fontsize=30)
plt.show()
In [41]:
# Distribution of Jersey Number
df['Jersey Number'].plot(kind = 'hist',bins= 320, color= 'blue', label = 'Jersey Number',
alpha = 1.0, grid = True, figsize = (10,5))
plt.legend
plt.xlabel('Number')
plt.ylabel('Players')
plt.title('Average Jersey Number')
plt.show()
In [42]:
# Distribution of Overall of players
df.Overall.plot(kind = 'hist',bins= 400, color= 'green',
label = 'Overall', alpha = 1.0, grid = False, figsize = (10,5))
plt.legend
plt.xlabel('Overall')
plt.ylabel('Players')
plt.title('Average Overall')
plt.show()
In [43]:
# Relation between Shotpower & Finishing
df.plot(kind = 'scatter', x='ShotPower', y='Finishing', alpha = 1.0, color = 'yellow')
plt.xlabel('ShotPower')
plt.ylabel('Finishing')
plt.title('ShotPower vs Finishing Scatter Plot')
Out[43]:
In [44]:
# Relation between Composure & Penalties
df.plot(kind = 'scatter', x = 'Composure', y = 'Penalties', alpha = 1.0, color = 'Orange')
plt.xlabel('Composure')
plt.ylabel('Penalties')
plt.title('Composure vs Penalties Scatter plot')
Out[44]:
In [45]:
# Relation between Ball Control and Short Passing
df.plot(kind = 'scatter', x = 'BallControl', y = 'ShortPassing',alpha = 1.0, color = 'Brown')
plt.xlabel('BallControl')
plt.ylabel('ShortPassing')
plt.title('BallControl vs ShortPassing Scatter Plot')
Out[45]:
In [46]:
# Relation between Sliding tackle & Interceptions
plt.figure(1,figsize=(10,5))
sns.regplot(df['SlidingTackle'],df['Interceptions'], color = 'Green')
plt.title('Sliding Tackle vs Interceptions Speed')
plt.show()
In [47]:
# Relation between Gk Diving & GK Positioning
plt.figure(figsize = (10,5))
sns.regplot(df['GKDiving'], df['GKPositioning'], color = 'Purple')
plt.title('GK Diving vs GK Positioning')
plt.show()
In [48]:
# The clubs and their players overalls
clubs = ('Juventus', 'Real Madrid', 'Paris Saint-Germain', 'FC Barcelona', 'Liverpool',
'Manchester United', 'FC Bayern München', 'Manchester City', 'Napoli')
plt.figure(figsize = (15,5))
df_club = df.loc[df['Club'].isin(clubs) & df['Age'] & df['Overall'] ]
ax = sns.barplot(x=df_club['Club'], y=df_club['Overall'], palette="rocket")
ax.set_title(label='Distribution overall in several clubs', fontsize=20);
In [49]:
#best players per each position with their age, club, and nationality based on their overall scores
df.iloc[df.groupby(df['Position'])['Overall'].idxmax()][['Position', 'Name', 'Age', 'Club', 'Nationality']]
Out[49]:
In [50]:
# Best features of players
pr_cols=['Crossing', 'Finishing', 'HeadingAccuracy', 'ShortPassing', 'Volleys',
'Dribbling', 'Curve', 'FKAccuracy', 'LongPassing', 'BallControl',
'Acceleration', 'SprintSpeed', 'Agility', 'Reactions', 'Balance',
'ShotPower', 'Jumping', 'Stamina', 'Strength', 'LongShots',
'Aggression', 'Interceptions', 'Positioning', 'Vision', 'Penalties',
'Composure', 'Marking', 'StandingTackle', 'SlidingTackle', 'GKDiving',
'GKHandling', 'GKKicking', 'GKPositioning', 'GKReflexes']
i=0
while i < len(pr_cols):
print('Best {0} : {1}'.format(pr_cols[i],df.loc[df[pr_cols[i]].idxmax()][1]))
i += 1
In [51]:
# Height of Players
plt.figure(figsize = (13, 8))
ax = sns.countplot(x = 'Height', data = df, palette = 'dark')
ax.set_title(label = 'Count of players on Basis of Height', fontsize = 20)
ax.set_xlabel(xlabel = 'Height in Foot per inch', fontsize = 16)
ax.set_ylabel(ylabel = 'Count', fontsize = 16)
plt.show()
In [52]:
# best players from each positions with their age, nationality, club based on their potential scores
df.iloc[df.groupby(df['Position'])['Potential'].idxmax()][['Position', 'Name', 'Age', 'Club', 'Nationality']]
Out[52]:
In [53]:
# Every Nations' Player and their overall scores
some_countries = ('England', 'Germany', 'Spain', 'Argentina', 'France', 'Brazil', 'Italy', 'Columbia')
df_countries = df.loc[df['Nationality'].isin(some_countries) & df['Overall']]
plt.figure(figsize = (10,5))
ax = sns.barplot(x = df_countries['Nationality'], y = df_countries['Overall'], palette = 'colorblind')
ax.set_xlabel(xlabel = 'Countries', fontsize = 15)
ax.set_ylabel(ylabel = 'Overall Scores', fontsize = 15)
ax.set_title(label = 'Distribution of overall scores of players from different countries', fontsize = 20)
plt.show()
In [54]:
some_clubs = ('Manchester United', 'Liverpool', 'Juventus', 'Napoli', 'Arsenal', 'Manchestar City',
'Tottenham Hotspur', 'FC Barcelona', 'Valencia CF', 'Chelsea', 'Real Madrid')
data_clubs = df.loc[df['Club'].isin(some_clubs) & df['Overall']]
plt.figure(figsize = (20,5))
ax = sns.barplot(x = data_clubs['Club'], y = data_clubs['Overall'], palette = 'deep')
ax.set_xlabel(xlabel = 'Some Popular Clubs', fontsize = 15)
ax.set_ylabel(ylabel = 'Overall Score', fontsize = 15)
ax.set_title(label = 'Distribution of Overall Score in Different popular Clubs', fontsize = 20)
plt.show()
In [55]:
# defining the features of players
player_features = ('Acceleration', 'Aggression', 'Agility',
'Balance', 'BallControl', 'Composure',
'Crossing', 'Dribbling', 'FKAccuracy',
'Finishing', 'GKDiving', 'GKHandling',
'GKKicking', 'GKPositioning', 'GKReflexes',
'HeadingAccuracy', 'Interceptions', 'Jumping',
'LongPassing', 'LongShots', 'Marking', 'Penalties')
# Top four features for every position in football
for i, val in df.groupby(df['Position'])[player_features].mean().iterrows():
print('Position {}: {}, {}, {}'.format(i, *tuple(val.nlargest(4).index)))
In [56]:
# Top five the most expensive clubs
df.groupby(['Club'])['Value'].sum().sort_values(ascending = False).head()
Out[56]:
In [57]:
# Top five the less expensive clubs
df.groupby(['Club'])['Value'].sum().sort_values().head()
Out[57]:
In [58]:
# Top five teams with the best players
df.groupby(['Club'])['Overall'].max().sort_values(ascending = False).head()
Out[58]:
In [59]:
# Relation between Age & Reactions
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['Reactions'], color = 'blue')
plt.title('Age vs Reactions')
plt.show()
In [60]:
# Relation between Age & Shotpower
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['ShotPower'], color = 'green')
plt.title('Age vs Shotpower')
plt.show()
In [61]:
# Relation between Age & Jumping
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['Jumping'], color = 'Orange')
plt.title('Age vs Jumping')
plt.show()
In [62]:
# Relation between Age & SprintSpeed
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['SprintSpeed'], color = 'Purple')
plt.title('Age vs SprintSpeed')
plt.show()
In [63]:
# Relation between Age & Stamina
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['Stamina'], color = 'Brown')
plt.title('Age vs Stamina')
plt.show()
In [64]:
# Relation between Age & Agility
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['Agility'], color = 'Yellow')
plt.title('Age vs Agility')
plt.show()
In [65]:
# Relation between Age & Strength
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['Strength'], color = 'aqua')
plt.title('Age vs Strength')
plt.show()
In [66]:
# Relation between Age & Vision
plt.figure(1,figsize=(10,5))
sns.regplot(df['Age'],df['Vision'], color = 'navy')
plt.title('Age vs Vision')
plt.show()
In [67]:
# Distribution of players according to their Overall
plt.figure(figsize = (20,10))
sns.countplot(df['Overall'], palette='rocket')
plt.show()
In [68]:
# Eldest players
df.sort_values(by = 'Age' , ascending = False)[['Name','Club','Nationality','Overall', 'Age' ]].head()
Out[68]:
In [69]:
# Youngest players
df.sort_values(by = 'Age' , ascending = True)[['Name','Club','Nationality','Overall', 'Age' ]].head()
Out[69]:
In [70]:
# Best Freekick takers
df.sort_values(by = 'FKAccuracy' , ascending = False)[['Name','Club','Nationality','Overall', 'Age','FKAccuracy']].head()
Out[70]:
In [71]:
# Best Penalty takers
df.sort_values(by = 'Penalties' , ascending = False)[['Name','Club','Nationality','Overall', 'Age','Penalties']].head()
Out[71]:
In [72]:
# Players with best ball control
df.sort_values(by = 'BallControl' , ascending = False)[['Name','Club','Nationality','Overall', 'Age','BallControl']].head()
Out[72]:
In [73]:
# Quick players
df.sort_values(by = 'SprintSpeed' , ascending = False)[['Name','Club','Nationality','Overall', 'Age','SprintSpeed']].head()
Out[73]:
In [74]:
# Age distrbution among famous clubs
clubs = ['Chelsea' , 'Arsenal', 'Juventus', 'Paris Sain-Germain' ,'FC Bayern München',
'Real Madrid' , 'FC Barcelona' , 'Borussia Dortmund' , 'Manchester United' ,
'FC Porto', 'Liverpool', 'Manchester City']
club_age = df.loc[df['Club'].isin(clubs) & df['Age']]
plt.figure(1 , figsize = (15 ,7))
sns.boxplot(x = 'Club' , y = 'Age' , data = club_age,palette='rocket')
plt.title('Age Distribution in famous clubs')
plt.xticks(rotation = 50)
plt.show()
In [75]:
# Overall Rating of the clubs
club_rating = df.loc[df['Club'].isin(clubs) & df['Overall']]
plt.figure(1 , figsize = (15 ,7))
sns.boxplot(x = 'Club' , y = 'Overall' , data = club_rating, palette='rocket')
plt.title('Overall Rating Distribution in famous clubs')
plt.xticks(rotation = 50)
plt.show()
In [76]:
# Best club
best_dict = {}
for club in df['Club'].unique():
overall_rating = df['Overall'][df['Club'] == club].sum()
best_dict[club] = overall_rating
best_club = pd.DataFrame.from_dict(best_dict,orient='index', columns = ['overall'])
best_club['club'] = best_club.index
best_club = best_club.sort_values(by = 'overall' , ascending = False)
plt.figure(1 , figsize = (15 , 6))
sns.barplot(x = 'club' , y = 'overall' , data = best_club.head(5),palette='rocket')
plt.xlabel("Club", size = 15)
plt.ylabel('Sum of Overall Rating of players in club', size = 15)
plt.title('Clubs with best Players (sum of overall ratings of players per club)', size = 25)
plt.ylim(2450 , 2600)
plt.show()
In [77]:
# Wage vs Overall
plt.figure(figsize=(25,10))
sns.barplot(data=df.head(30),y='Wage',x='Overall',palette='rocket')
plt.title('Wage vs Overall',size=35)
plt.xlabel("Overall",size=25)
plt.ylabel('Wage',size=25)
plt.show()
In [78]:
# Wage vs Potental
plt.figure(figsize=(25,10))
sns.barplot(data=df.head(30),y='Wage',x='Potential',palette='rocket')
plt.title('Wage vs Potential',size=35)
plt.xlabel("Wage",size=25)
plt.ylabel('Potential',size=25)
plt.show()
This gives us a mere idea that wages of players depends more on their overall performance than their potential
In [79]:
# Count of players for each position
df3=pd.DataFrame(df['Position'].value_counts())
df3.reset_index(inplace=True)
df3.rename(columns={'index':'Position',"Position":"Count"},inplace=True)
df3.head(10)
Out[79]:
In [80]:
plt.figure(figsize=(20,10))
sns.countplot(x='Position',data=df)
plt.title("Player count per position",size=25)
plt.xlabel("Position",size=20)
plt.ylabel("Count",size=20)
plt.show()
print("Maximum players were found with position {} with a count of {}".format(df3['Position'][0],df3['Count'][0]))
In [81]:
import plotly.graph_objects as go
fig=go.Figure(data=go.Scatterpolar(r=df3['Count'].head(),theta=['ST','GK','CB', 'CM','LB'],fill='toself'))
fig.update_layout(polar=dict(radialaxis=dict(visible=True),),showlegend=False)
fig.show()
In [82]:
#Players whose contract has ended (till 2019)
df2=df[df['Contract Valid Until']<='2019'][['ID','Name','Age','Nationality','Contract Valid Until','Position','Club']]
print("Total Contracts ended:",df2.shape[0])
df2[['ID','Name','Age','Nationality','Position','Club']].head(15)
Out[82]:
In [83]:
# Players' age when their contract ended (2019)
plt.figure(figsize=(20,10))
sns.barplot(data=df2.head(15),x=df2['Name'].head(15),y=df2['Age'].head(15))
plt.title('Age of players when their contract ended (top 15)',size=20)
plt.xlabel("Names",size=25)
plt.ylabel('Age',size=25)
plt.show()
print("Average age of players whose contract ended:",int(df2['Age'].mean()))
In [84]:
#Contracts ended per year
plt.figure(figsize=(5,5))
sns.countplot(x=df2['Contract Valid Until'])
plt.title('No. of contracts ended in previous years',size=15)
plt.xlabel("Years",size=15)
plt.ylabel('Count',size=15)
plt.show()
print("Contracts ended in 2019:",df2['Contract Valid Until'].value_counts()[0])
print("Contracts ended in 2018:",df2['Contract Valid Until'].value_counts()[1])
In [85]:
# Contracts ended for various positions
d=pd.DataFrame(df2['Position'].value_counts())
d.reset_index(inplace=True)
d.rename(columns={"index":"Position","Position":"Count"},inplace=True)
d.head()
Out[85]:
In [86]:
plt.figure(figsize=(20,10))
sns.countplot(x=df2['Position'])
plt.title('No. of contracts ended per position',size=20)
plt.xlabel("Years",size=15)
plt.ylabel('Count',size=15)
plt.show()
print("Most contracts ended for position {} and their count is: {}".format(d['Position'][0],d['Count'][0]))
In [87]:
#Contracts ended per club
d1=pd.DataFrame(df2['Club'].value_counts())
d1.reset_index(inplace=True)
d1.rename(columns={"index":"Club","Club":"Count"},inplace=True)
d1.head(10)
Out[87]:
In [88]:
plt.figure(figsize=(15,10))
plt.bar('Club','Count',data=d1.head(10),width=0.7,color='red')
plt.title('No. of contracts ended for clubs (top 10)',size=20)
plt.xlabel("Clubs",size=25)
plt.xticks(rotation = 90,fontsize=15, fontname='sans-serif')
plt.ylabel('Count',size=25)
plt.show()
print("\nMost contracts ended for club {} and its count is: {}".format(d1['Club'][0],d1['Count'][0]))
In [89]:
df1=pd.DataFrame(df['Nationality'].value_counts())
df1.reset_index(inplace=True)
df1.rename(columns={'index':"Nation","Nationality":"Count"},inplace=True)
df1.head()
Out[89]:
In [90]:
# Most players from a country
plt.figure(figsize=(25,10))
sns.barplot(data=df1.head(20),x='Nation',y='Count',palette='rocket')
plt.title('Most players from a country (top 20)',size=35)
plt.xlabel("Nations",size=25)
plt.ylabel('Count',size=25)
plt.show()
In [91]:
# Pie chart depiction of countries with most players
plt.figure()
fig1,ax1=plt.subplots(figsize=(8,8))
plt.subplots_adjust(left=0.5,wspace=0.2)
ax1.pie(df1['Count'].head(7),explode=(0.1,0,0,0,0,0,0,),labels=df1['Nation'].head(7),autopct='%1.1f%%',shadow=True,startangle=90)
ax1.axis('equal')
plt.legend(df1['Nation'].head(7),loc="best")
plt.tight_layout()
plt.show()
This signifies that England and Germany have most footballers.
In [92]:
import folium
from geopy.geocoders import Nominatim
import requests
In [93]:
lat=[]
lng=[]
geolocator=Nominatim(user_agent='foursquare_api')
for nm in df1['Nation'].head(20):
location=geolocator.geocode(str(nm),timeout=10)
lat.append(location.latitude)
lng.append(location.longitude)
In [94]:
for_map=pd.DataFrame(df1.head(20))
for_map['Latitude']=lat
for_map['Longitude']=lng
for_map.head()
Out[94]:
In [95]:
wm=folium.Map(zoom_start=2,location=[0,0])
mp=folium.map.FeatureGroup()
for i,j,k in zip(for_map['Latitude'],for_map['Longitude'],for_map['Count']):
mp.add_child(folium.CircleMarker(location=[i,j],radius=5,color='red',fill_color='Yellow'))
folium.Marker([i,j],popup='Player Count '+str(k)).add_to(mp)
wm.add_child(mp)
wm
Out[95]:
Data Modelling
1. Regression
1.1 To find 'Overall' of footballers
Here, we select some attributes of the players to predict the 'Overall' so as to get the total performance measure of the players.
In [96]:
predictors=df[['Overall','Potential','Value','Wage','Skill Moves',
'Crossing','Finishing','HeadingAccuracy','ShortPassing','Volleys','Dribbling',
'Curve','FKAccuracy','LongPassing','BallControl','Acceleration','SprintSpeed',
'Agility','Reactions','Balance','ShotPower','Jumping','Stamina','Strength',
'LongShots', 'Aggression', 'Interceptions', 'Positioning', 'Vision',
'Penalties', 'Composure', 'Marking', 'StandingTackle', 'SlidingTackle',
'GKDiving', 'GKHandling', 'GKKicking', 'GKPositioning', 'GKReflexes']]
predictors.head()
Out[96]:
As seen from the scatter and regression plots above, the predictor's columns vary linearly with 'Overall'. Hence we select the 'Multiple Linear Regression' algorithm.
Importing needed modules
In [97]:
from sklearn import linear_model
from sklearn.metrics import r2_score
Creating training and testing datasets
In [98]:
ms=np.random.rand(len(df))<0.75
train=predictors[ms]
test=predictors[~ms]
In [99]:
x=np.asanyarray(train[['Potential','Value','Wage','Skill Moves','Crossing',
'Finishing','HeadingAccuracy','ShortPassing','Volleys','Dribbling',
'Curve','FKAccuracy','LongPassing','BallControl','Acceleration','SprintSpeed',
'Agility','Reactions','Balance','ShotPower','Jumping','Stamina','Strength',
'LongShots', 'Aggression', 'Interceptions', 'Positioning', 'Vision',
'Penalties', 'Composure', 'Marking', 'StandingTackle', 'SlidingTackle',
'GKDiving', 'GKHandling', 'GKKicking', 'GKPositioning', 'GKReflexes']])
y=np.asanyarray(train['Overall'])
In [100]:
x
Out[100]:
In [101]:
y
Out[101]:
Fitting the model
In [102]:
regr=linear_model.LinearRegression()
regr.fit(x,y)
Out[102]:
In [103]:
print("Coefficients:",regr.coef_)
print("\nIntercept:",regr.intercept_)
Prediction
In [104]:
Y=regr.predict(test[['Potential','Value','Wage','Skill Moves','Crossing',
'Finishing','HeadingAccuracy','ShortPassing','Volleys','Dribbling',
'Curve','FKAccuracy','LongPassing','BallControl','Acceleration','SprintSpeed',
'Agility','Reactions','Balance','ShotPower','Jumping','Stamina','Strength',
'LongShots', 'Aggression', 'Interceptions', 'Positioning', 'Vision',
'Penalties', 'Composure', 'Marking', 'StandingTackle', 'SlidingTackle',
'GKDiving', 'GKHandling', 'GKKicking', 'GKPositioning', 'GKReflexes']])
Comparison
In [105]:
x1=np.asanyarray(test[['Potential','Value','Wage','Skill Moves','Crossing',
'Finishing','HeadingAccuracy','ShortPassing','Volleys','Dribbling',
'Curve','FKAccuracy','LongPassing','BallControl','Acceleration','SprintSpeed',
'Agility','Reactions','Balance','ShotPower','Jumping','Stamina','Strength',
'LongShots', 'Aggression', 'Interceptions', 'Positioning', 'Vision',
'Penalties', 'Composure', 'Marking', 'StandingTackle', 'SlidingTackle',
'GKDiving', 'GKHandling', 'GKKicking', 'GKPositioning', 'GKReflexes']])
y1=np.asanyarray(test['Overall'])
In [106]:
a={"Predicted Values":Y,"Actual Values":y1}
comp=pd.DataFrame(a)
comp=comp.astype({"Predicted Values":'int64'})
comp.set_index(["Predicted Values","Actual Values"],inplace=True)
comp.head()
Out[106]:
Model Evaluation
Accuracy Measurement
In [107]:
f1=np.mean((Y-y1)**2)
print("Residual sum of squares: {}".format(f1))
In [108]:
f2=regr.score(x1,y1)
print("Variance Score:",f2)
In [109]:
f3=np.mean(np.absolute(Y-y1))
print("Mean absolute error:",f3)
In [110]:
f4=r2_score(Y,y1)
print("R2-Score:",f4)
print("R2-Score Percentage",round(f4*100,3),"%")
In [112]:
from sklearn import metrics
f5=np.sqrt(metrics.mean_squared_error(Y,y1))
print("MSE:",f5)
Model Summary
In [113]:
a={"Feature":["Residual Sum Of Squares","Variance Score","Mean Absolute Error","Mean Squared Error","R2-Score Percentage"],
"Value":[round(f1,3),round(f2,3),round(f3,3),round(f5,3),round(f4*100,3)]}
summary=pd.DataFrame(a)
summary.set_index(['Feature','Value'],inplace=True)
summary
Out[113]:
2. Clustering
Here, we cluster the players as per their potenial and overall performance, so as to assign them with certain labels.
We will use the k-means clustering algorithm.
In [114]:
from sklearn.cluster import KMeans
We tend to cluster the players into 3 major categories:
- Best
- Good
- Average
- Below Average
In [115]:
predictors1=df[['Overall','Potential']]
Data Normalization/Standardization
In [116]:
from sklearn.preprocessing import StandardScaler
X=predictors1.values[:,1:]
clus_dataset=StandardScaler().fit_transform(X)
In [117]:
# Original Values
X
Out[117]:
In [118]:
# Normalized Values
clus_dataset
Out[118]:
Fitting the Model
In [119]:
kMeans=KMeans(init='k-means++',n_clusters=4)
kMeans.fit(clus_dataset)
Out[119]:
In [120]:
# Generating labels
labels=kMeans.labels_
In [121]:
predictors1['Labels']=labels
df['Labels']=labels
Cluster 1
In [122]:
cluster1=df.loc[df['Labels']==0,df.columns[[1]+[4]+[5]+[82]]]
pos1=cluster1['Potential'].mean()
ov1=cluster1['Overall'].mean()
mem1=cluster1.shape[0]
print("Mean Potential of cluster:",pos1)
print("Mean Overall of cluster:",ov1)
print("Members in cluster 1:",mem1)
cluster1.head()
Out[122]:
Cluster 2
In [123]:
cluster2=df.loc[df['Labels']==1,df.columns[[1]+[4]+[5]+[82]]]
pos2=cluster2['Potential'].mean()
ov2=cluster2['Overall'].mean()
mem2=cluster2.shape[0]
print("Mean Potential of cluster:",pos2)
print("Mean Overall of cluster:",ov2)
print("Members in cluster 1:",mem2)
cluster2.head()
Out[123]:
Cluster 3
In [124]:
cluster3=df.loc[df['Labels']==2,df.columns[[1]+[4]+[5]+[82]]]
pos3=cluster3['Potential'].mean()
ov3=cluster3['Overall'].mean()
mem3=cluster3.shape[0]
print("Mean Potential of cluster:",pos3)
print("Mean Overall of cluster:",ov3)
print("Members in cluster 1:",mem3)
cluster3.head()
Out[124]:
Cluster 4
In [125]:
cluster4=df.loc[df['Labels']==3,df.columns[[1]+[4]+[5]+[82]]]
pos4=cluster4['Potential'].mean()
ov4=cluster4['Overall'].mean()
mem4=cluster4.shape[0]
print("Mean Potential of cluster:",pos4)
print("Mean Overall of cluster:",ov4)
print("Members in cluster 1:",mem4)
cluster4.head()
Out[125]:
As per the results, we assign the following tags to the clusters:
- Cluster 1: Best
- Cluster 2: Below Average
- Cluster 3: Good
- Cluster 4: Average
Summary
In [128]:
clus={"Index":[1,2,3,4],"Tags":["Best","Good","Average","Below Average"],
"Potential Mean":[round(pos1,3),round(pos3,3),round(pos4,3),round(pos2,3)],
"Overall Mean":[round(ov1,3),round(ov3,3),round(ov4,3),round(ov2,3)],
"Players":[mem1,mem3,mem4,mem2],"Cluster":[3,2,4,1]}
clus_sum=pd.DataFrame(clus)
clus_sum.set_index(['Index'],inplace=True)
clus_sum
Out[128]:
Visualizing Clusters
In [129]:
sns.set_style('whitegrid')
sns.lmplot('Overall','Potential',data=df,hue='Labels',palette="husl",size=6,aspect=1,fit_reg=False)
Out[129]:
